While the other answers explain very well the issues that stealth would have to overcome, I'm going to have to take a slightly different tack and say **it is possible**.

**Heat**: Okay, purely using heatsinks seems to be out, unless the ship is very large.  However, it is possible to collimate the IR output of a ship, and not by using a laser either (which would seem to violate thermodynamics).  Instead, simply have a hot core which is exposed from just one narrow opening/tunnel/exhaust port.  Anyone looking from the direction of the beam will see it; but anyone slightly off-axis won't. IR lenses are difficult, but IR mirrors are quite easy.  Actively cool the port (or baffles along the edges of the beam), and also the outside of the ship.  Use a blob of material that heat is actively dumped into (for example with peltier effect or similar), at the focus of a system of mirrors which collects and collimates the emitted IR.  I think the IR can be collimated to about the same degree as a laser, and thus the odds of it hitting a sensor platform are quite low.  This is not perfect stealth, but it is stealth for all practical purposes.

**Drive**: The easiest stealth drive is something like a railgun, only used for propulsion.  It can produce cold exhaust chunks (macroscopic particles, not gas - if needed, pre-chilled), and any residual IR would be hard to see because of the velocity.  I don't know why everyone here is thinking *hot* exhaust... 

**Occlusions**: Okay, this is a tough one.  Aside from some handwaving about metamaterials, I'd say the practical approach is to stick to a natural object that is travelling in the right direction.  The exact size of a small object ought to be much harder to gauge from occlusions than whether it is there at all.  Thus, practical stealth would involve brief runs in free space between convenient camouflage asteroids.